Implement
AI automation for Swiss SMEs
Workflows that run instead of being planned. n8n as orchestrator, ChatGPT, Claude or Copilot as the intelligence layer, Custom GPTs for knowledge apps.
- n8n self-hosted or Cloud
- Custom GPT, Claude Projects, Copilot
- revDPA-compliant configuration
- On-prem capable (ThinkStation PGX)
- CHF 2,200 / day
You have routine tasks that AI could handle: classifying mail, filling quote templates, reading in documents, generating standard replies.
But you have no appetite for a six-month implementation with three consultants and a steering committee. Waldsee Consulting builds these automations with n8n as the orchestrator and an LLM layer (ChatGPT Enterprise, Claude, Copilot, or on-prem). Pragmatic, fast, at the day rate CHF 2,200. First working version typically after 3–7 days.
Three classes of automation
| Class | Examples | Stack | Effort |
|---|---|---|---|
| Mail / Office | Inbox triage, reply suggestions, template filling | n8n + IMAP/SMTP + LLM | 5–10 days |
| Document workflows | OCR + classification + DMS filing, contract analysis | n8n + DMS API + LLM (RAG if needed) | 8–15 days |
| Knowledge apps | Employee bot over internal docs, FAQ bot over Confluence/Notion | Custom GPT or Claude Project + RAG index | 5–12 days |
Stack overview
- Orchestration: n8n (self-hosted preferred for data control, Cloud possible)
- LLM layer: ChatGPT Enterprise, Claude, Microsoft Copilot, or on-prem on Ollama
- Knowledge indexes: RAG pipelines with vector DBs (Qdrant, Postgres+pgvector)
- Interfaces: email (IMAP/SMTP), CRM, DMS, Slack, Teams
- For sensitive data: on-prem setup on Lenovo ThinkStation PGX
ChatGPT Enterprise, Copilot, or n8n + custom?
| You use… | Recommendation |
|---|---|
| Microsoft 365 day to day | Copilot M365 as default + n8n for special workflows |
| Google Workspace | ChatGPT Enterprise + n8n |
| Open-source-minded / IP-sensitive | n8n + on-prem LLM (Ollama) |
| Mix | Hybrid, we decide in the architecture conversation |
Common questions
What is n8n?
An open-source workflow tool, similar to Zapier or Make.com, but self-hostable. We use n8n as the orchestration layer because your data stays under your control and there are no per-run licence costs.
What does n8n cost?
n8n Community Edition: free, self-hosted. n8n Cloud: from USD 20/month. Hosting effort when self-hosted: minimal, a small VPS is enough for most SMEs.
ChatGPT Enterprise or Copilot?
It depends on your stack. M365 shops: Copilot. Google Workspace shops: ChatGPT Enterprise. Both under revDPA only with a DPA contract. For IP-sensitive data: on-prem instead of either.
What does a typical automation cost?
5–15 days of Waldsee effort at the day rate CHF 2,200, plus tool licences if needed.
How fast does the first version run?
Typically 3–7 days until a first working workflow is in place. Refinement after that is iterative.
What if the automation makes a mess?
Every automation is built with logging, error mail alerts and a manual override option. For critical workflows: human in the loop for final approval.